In January, I deleted Twitter from my phone. The app was a charming distraction, I decided, and 2015 would be a year of productivity upgrades. Three weeks later, I cheated by opening Twitter on my iPhone browser. Four weeks later, I was refreshing the same page every other hour. Five weeks later, my productivity experiment in tatters, I re-downloaded the app.

This is when I noticed a new feature: View Tweet Activity. It is an engagement "dashboard"—that is, numbers with pictures—that tells you how many times your tweet appeared on users' glassy screens, how many times they clicked it, and how many times they shared it.

These are the metrics that journalists and marketers crave, because they can illustrate corporate PowerPoints and answer valuable quantitative questions (e.g. What phrases best correlate with Twitter engagement for our brand?). But the numbers also get at the qualitative questions that anybody living publicly online is wondering quietly, like Was my joke any good? and, even more intimately, Do people care what I have to say?

Last Monday, I published an article about the history of American innovation as seen through a study of patent text literature. This study found that chemistry concepts dominated science in the early 20th century, but from the 1980s on, the most-cited terms in patent texts were almost entirely in the fields of medicine and computers. Yesterday, chemistry; today, computers. This seemed like a catchy parallel, which might strike some as illuminating and others as over-simplifying. In other words, the perfect tweet.

I wrote this message, with a link, and a picture:

By Friday morning, it had about 155,260 impressions. According to the new Tweet activity dashboard, 2.9 percent of those users clicked the image, and 1.1 percent retweeted or favored it ...

... but just 1 percent clicked on the link to actually read my story. One percent.

Even worse, of the 9,017 people who clicked somewhere, anywhere on my message, just one in six of those clicks actually went to The Atlantic website. Quantitatively speaking, my viral tweet had the click-through rate of a digital display ad in East Asia.

Every good media organization knows that the road to traffic leads through Facebook rather than Twitter. Even so, I thought the sharing economy of the Internet shared a bit more than this.

A tweet with 10,000 interactions is an exception, and I was interested in the rule. So I went to Twitter's user analytics page to download the data on my 100 most popular tweets of the last year. If I could prove to my bosses (and to myself) that Twitter could, even occasionally, deliver meaningful audiences, it might validate my infatuation. Alas, my most popular tweets averaged a click-through rate of about 1.7 percent, still quite near the rate of conversions on flash-media East Asian display ads. Without revealing numbers that will get me in trouble with my bosses, I concluded that my prodigious use of Twitter in the last 30 days has cumulatively driven less traffic to than one of my below-average stories.

Is the social web just a matrix of empty shares, of hollow generosity? As Chartbeat CEO Tony Haile once said (on Twitter), there is "effectively no correlation between social shares and people actually reading." People read without sharing, but just as often, perhaps, they share without reading.

When I graphed my 100 most popular tweets by clicks (graph one below) and engagement (graph two), the result is a jagged mess. Many of my most engaged tweets barely generated any clicks. Readers treated the URL in the tweet as a footnote. Some tweets, you might say, are "too good to click": They offer such a complete story that it leaves no curious itch.

Twitter as Television: Watch But Don't Click

There used to be a vague sense that Twitter drives traffic, and traffic drives renown (or fame, or pride, or whatever word defines the psychic benefit of public recognition). Instead, the truth is that Twitter can drive one sort of renown (there are some people who are Twitter-famous), and traffic affords a different psychic currency. But they are nearly independent variables.

I wanted another perspective. So I wrote an email to my Atlantic colleague Robinson Meyer, who has established himself as one of the smarter commentators on the peculiarities of digital media. I told him I had created something that 150,000 people had seen, 9,000 people had interacted with, and just 1,500 had followed to our site to actually read. (So, 99 percent of my labor on Twitter went to Twitter, and 1 percent went to The Atlantic. That's not a very good deal for our boss!) It was, I told Rob, as if Twitter had built a feature illustrating how much people were ignoring its power users.

Rob responded:

It’s kind of a funny feature, right? It can tell you exactly how many people saw and how many people ignored it, basically (minus the impossible metric: how many people thought of you.) It’s a funny bit of information to give away.

Maybe there’s a comparison to Snapchat Discover here. Twitter was a great mobile journalism platform, and it was very good at getting people to read… Twitter. But people functionally left it no more than they actually can leave Snapchat. So at least Snapchat lets you take some reader ad revenue.

(Rob’s response alluded to Snapchat’s new “Discover” feature, which serves up content from cable networks and popular news sites within Snapchat.)

It's fair to come away from these metrics thinking that Twitter is worthless. But that's an unsophisticated conclusion. The more sophisticated takeaway is that Twitter is worthless for the limited purpose of driving traffic to your website, because Twitter is not a portal for outbound links, but rather a homepage for self-contained pictures and observations. (The irony is that the more journalists consider Twitter a portal, the better Twitter becomes as a home for other people to stay, including other journalists.)

Two weeks ago, John Herrman observed that as readers' digital attention scatters to Snapchat, Twitter, Instagram, and Facebook, publishers will acknowledge that their websites are anachronisms. It's hard to say that this is happening today. Most major websites are seeing growing traffic. But there are only so many hours one can look at a screen in a day. More of those hours are going to mobile devices. A growing share of those hours (and their corresponding dollars) are going to communications apps, like Twitter. And, by my calculations here, Twitter is sending less than 2 percent of its overall engagement back to the web. Apps don't pay my rent. A website does.

In the last month, I've created nearly 2 million impressions for Twitter. Whether that is good for my Twitter persona and my pride is a qualitative question whose answer resides outside the bounds of an analytics dashboard. But it is quantitatively not a good deal for The Atlantic. Something I already suspected has now been made crystal clear: 99 percent of my work on Twitter belongs to Twitter.